agent-readylisted

# agent-ready Make a website **readable by AI agents** and **prove** it. Every score comes from the deterministic core (the `agent-visible` CLI), never from your own judgment — this is the anti-hallucination guarantee. **You reason and fix; the core measures.** AI answer engines like ChatGPT and Perplexity crawl with bots (GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot) that **do not run JavaScript** — Google/Gemini's renderer (WRS) is the notable exception. If a site renders content client-side, those non-JS bots see an empty shell and the site is invisible in their answers. agent-ready measures what those bots actually receive and closes the gap. ## Prerequisite Install the scanner once so `agent-visible` is on PATH (run in the agent-ready repo): `npm install && npm run build && npm link` Then invoke it anywhere as `agent-visible <target> --json`. (Without `npm link`, run it from the repo as `node dist/cli.js <target> --json`.) ## The rubric (v2 — total 100, transparent in `src/rubric.ts`) A five-tier funnel mirroring how an AI crawler reads a site: | Tier | Pts | Checks | |------|----:|--------| | **Access** | 15 | reachable (HTTP 200); not blocked/cloaked for AI user-agents | | **Permission** | 15 | robots.txt valid; AI bots not Disallowed; Content-Signal declared | | **Readability** | 35 | real text in raw HTML; not an empty SPA shell; headings; semantic landmarks | | **Understandability** | 25 | schema.org JSON-LD; title; meta description; Open Graph; `lang`; ca